I am Shadrach Chinecherem Eze. I hold a bachelor’s in Pharmacy and a master’s degree in Drug discovery. My focus in this space is to discuss scholarship, academics, excellence, purpose and to provide inspiration to students Let’s follow each other 😁

🏭 INVEST IN ENUGU! 💚 Another milestone for Enugu State! 🚀 Watch the official commissioning of Anchorage Pharmaceutical Industry Limited by Governor Peter Ndubuisi Mbah. Enugu is opening its doors to investment, industrial growth, job creation and economic opportunities.🇳🇬🏗️

If you want to know which countries will drive tomorrow’s scientific and political advancements, watch this detailed report on the 2025 PISA rankings. Unfortunately, Nigeria is nowhere to be found. Even more concerning, the bottom four countries are all African nations. The geography of talent today shapes the geography of industrial advancement tomorrow. This is precisely why I want South East Nigeria to be assessed and ranked alongside the rest of the world, just like Shenzhen in China. We cannot aspire to compete globally if we are unwilling to measure ourselves against global standards.

Yann LeCun, one of the world’s foremost AI, machine-learning and robotics researchers, an engineer, NYU professor and ACM Turing Award laureate, has been torn to pieces today by Elon Musk fans for daring to inject some reality into the Tesla Roadster debate. The objection isn’t that the technology isn’t impressive. It is how these capabilities are being marketed, with deposits being taken on the basis of what we’ve been shown. The physics remain the problem. A production car carrying enough compressed air to deliver those levels of thrust would face enormous compromises in weight, storage capacity and recharge requirements. At best, you’re looking at seconds of extreme thrust, followed by specialist recharging that can absolutely not be integrated into the car - literally impossible because of the weighty equipment needed & the energy required for that equipment to compress 600L of air for a few seconds of thrust. And let’s not forget that the Roadster 2.0 was announced in 2017. Nine years later, instead of finally putting that car on the road, we’re watching a new version of the same old story. This isn’t hate. It’s realism based on experience, engineering and physics. And when someone like @ylecun makes essentially the same observations, the response shouldn’t be “What have you ever achieved?” or "what do you know?" - sure, use that one on me, but not on Yann.

tragic

We undertook our economic reforms to build a stronger foundation for investment and give more Nigerians the opportunity to earn a better living. I welcome the World Bank Group’s work to mobilise up to $5.79 billion in private financing for Nigeria. This partnership supports our efforts to help businesses expand and attract investment into agriculture and digital infrastructure. Nigeria has gone a step further to put together a pipeline of 65 bankable projects seeking $78.5 billion in investment. Our Minister of Finance and Coordinating Minister of the Economy, Taiwo Oyedele @taiwoyedele, presented these opportunities alongside CBN Governor Olayemi Cardoso at the Nigeria Investors Forum, on the sidelines of the 2026 IMF–World Bank Annual Meetings in Bangkok. I invite Nigerian and international investors to examine these projects and engage directly with their sponsors through Projects Nigeria. Our administration will work with you to resolve obstacles and move viable investments into execution, creating jobs and expanding opportunities for our people. Explore the pipeline at finance.gov.ng/projectsniger… Bola Ahmed Tinubu, GCFR President @NGRPresident @NigeriaGov .@FinMinNigeria .@cenbank @WorldBankAfrica @WorldBankGroup

马斯克这次直接把 Roadster 发布会开成了科幻电影！😱 完整版来了，帮大家整理了几个高能时间点： 03:15 首次飞行演示 05:38 飞行汽车成真 06:02 风扇吸力系统，汽车倒挂 07:52 SpaceX 火箭推进器，空中特技 08:52 汽车竟然玩起了 Kickflip 14:21 极限加速演示 16:22 0-60 英里加速仅需 0.81 秒 19:04 计划 2028 年开始量产 建议从头看到尾，太炸裂了！

UC Berkeley's "CS 188: Introduction to Artificial Intelligence" by John Canny & Oliver Grillmeyer The legendary Pacman-based AI course: search, CSPs, games, MDPs & ML, taken by tens of thousands online. Course Website: inst.eecs.berkeley.edu/~cs18… Lecture Videos: piped.video/playlist?list=PL…

We had an online seminar recently. The speaker mentioned an open problem at the end of the talk, which would be a natural continuation of the published work. It seemed like a good problem. It asks whether or not a certain construction can exist. Well defined, natural, nice. I asked Astra. It solved the problem after 2-3 prompts. The first solution (a concrete example) seems too simple, so I asked Astra to find something more complicated. Now there is a second example. I asked Astra to explain to me why the solution works, this happened on the way back home from work. Travel on the tram and the metro used efficiently. At home I also asked Astra to compile a research article. The end result is a good starting point for a paper. Of course it looks like AI slop, but it is way less slop then what earlier models would produce. It would take only a few days of concentrated work to make an actual journal quality paper out of it. All of this feels... illegal, pointless, and exciting at the same time. So many conflicting emotions. What do we gain from all of this? Should we publish papers? Or just shout into the world that we know an answer? I am really not sure. Naturally, I informed the speaker, he was interested. I sent him the document. I am waiting for the answer. Minimal efforts, some understanding, and now a new result that would seem non-trivial for most of my colleagues. A new paper that had been welcome and appreciated at any time pre-AI. Where do we really go from here...

今夏公開されたフランス映画「ド・ゴールの戦い」(2026)の予告編。2部構成で第一部が「抵抗」、第二部が「解放」、全編で5時間を超える超大作。尺の長さからいくと日本での劇場公開は難しいかも

Study math. Study math to think. Study until complex systems stop feeling random. Study the mental models behind the formulas. Study for the person who thought they weren't smart enough in school. A simple habit of rigorous thinking will change absolutely everything.

This semester I am teaching a new course on Agentic Systems, which explores the full stack of AI systems: from building and optimizing agent harnesses to serving and optimizing LLM serving infrastructure. Students learn by building, measuring, and optimizing real agents and systems, balancing capability, latency, and cost. Our goal is not just to teach today's AI technologies, but to prepare students to build tomorrow's. All lectures, readings, and assignments are publicly available at cs2680.seas.harvard.edu

My latest video is out now! Here's all basic graph theory concepts, visually explained, in 45 minutes: piped.video/meGET4i3QMc?si=omAm…

Africa’s creative industry is blooming. 🇳🇬 At Nerdwork+, we’re building a home for comics, novels, animation, and games. With releases like XION and JAJA: Lost Souls, we’re championing Nigerian culture and scaling African IPs. Discover our worlds 👇🏾 nerdworkplus.com/

Want to learn about venture capital? Watch my course. lnkd.in/p/eadn8urr

Here are the Playlists for the courses that I have taught: 1. Mathematical Foundations of Machine Learning - piped.video/playlist?list=PL… 2. Mathematical Foundations of Generative Models - piped.video/playlist?list=PL… 3. Mathematical Foundations of Genrative AI - piped.video/playlist?list=PL…

Yann LeCun's latest talk at ETH Zürich "You should not work on LLM. At least if you're in academia, you should absolutely not work on LLMs. There is nothing you can bring to the table. If you're interested in making real progress in AI, in sort of grounded AI for the real world, if you want physical AI, don't work on LLMs, and don't work on generative models either. So, as you can probably guess, this does not make me very popular in Silicon Valley." ---- From "Perfology Clips" YouTube channel, (link in comment)

The PhD thesis of my 15th PhD student, Shibhansh Dohare (@ShibhanshD), is now available. Title: Learning Forever using Artificial Neural Networks Url: incompleteideas.net/papers/D… Abstract: In this dissertation, I study the ability of artificial neural networks to learn new things. I refer to this ability as plasticity. Plasticity is a desirable property of intelligent systems in the real world. Systems that interact with the real world can become outdated if they do not learn from the new information revealed to them by the extremely large and partially observable real world. All naturally intelligent systems continually learn new things to keep up with the changes in the world. Many applications like predicting markets or weather, learning human preferences, controlling factories or prosthetics, etc require AI systems that continually learn new things. I provide direct and thorough demonstrations of loss of plasticity in deep learning systems. The demonstrations range from simple feed-forward networks to deep residual networks and cover a wide range of algorithmic choices. While seeking an understanding of the phenomenon, I found that loss of plasticity occurred simultaneously with hidden units becoming dormant and becoming similar to each other, hampering the ability of the networks to learn new things. Next, I describe the continual backpropagation algorithm. I designed this algorithm, in collaboration with others, to maintain plasticity in artificial neural networks. Continual backpropagation is a simple extension of the conventional backpropagation algorithm. The conventional backpropagation has two main parts: initialization with small random weights and gradient descent at each step. In addition to gradient descent, continual backpropagation reinitializes a small fraction of units at each step. Continual backpropagation indefinitely maintained plasticity in many continual supervised learning problems. Finally, I study plasticity of deep reinforcement learning systems. First, I show that plasticity loss has an effect even in stationary reinforcement learning problems. It is unsurprising that plasticity loss affects stationary reinforcement learning because it requires continual learning due to changing policies and bootstrapping. This effect is visible under long-term learning. I found that some commonly used RL algorithms do not scale with data. Their performance drops as they continue interacting with the environment. Plasticity loss hampers the ability of these algorithms to recover and relearn when the performance drops, resulting in agents that do not get better with experience. Continual backpropagation allows these algorithms to recover when performance drops, leading to agents whose performance keeps improving with experience. Second, I show that loss of plasticity is a significant problem in continual reinforcement learning problems, and continual backpropagation maintains plasticity in these problems. Shibhansh is now a research scientist at Keen Technologies.

See why history is good to know…. Happened inside a Lagos, Nigeria courthouse in 1938… Two Germans begging not to be sent home.

Il y a quelques années, j'ai vu une vidéo où Elon Musk expliquait que la majorité des PhD ne servaient à rien. Sur le moment, j'ai trouvé ça excessif. Puis j'ai commencé à bosser dans l'IA. Et aujourd'hui, je lis des papiers de recherche toutes les semaines. Et bien souvent, je referme le PDF en me disant : mais putain, à quoi ça sert ? Des mois de travail, souvent payés par l'argent public, pour produire des trucs qui ne créent absolument aucune valeur. Aucun produit. Aucune avancée. Aucune idée nouvelle. De la branlette intellectuelle. Et j'ai fini par comprendre. Quand un système est gangrené par l'idéologie, il cesse de récompenser ce qui marche. Il récompense ce qui se conforme. On ne cherche plus la vérité, on cherche la publication. On ne cherche plus à construire, on cherche à être cité par ses pairs. Le réel n'est plus le juge. Le comité l'est. C'est pour ça que ce que fait Alex Karp chez Palantir est génial. Il ne se plaint pas de l'université. Il la contourne. Il prend des jeunes de 18 ans, les met sur de vrais problèmes, avec de vrais enjeux, et les juge sur une seule chose : est-ce que ça marche ? C'est la plus vieille pédagogie du monde. Et c'est la plus efficace. Mais Palantir ne peut pas tout faire. C'est l'intégralité du modèle éducatif qu'il faut reset. De la maternelle jusqu'au doctorat. Regardez ce qu'a fait Elon Musk avec Ad Astra, l'école qu'il a créée pour ses enfants. Pas de classes par âge. Pas de notes. Pas de cours magistraux sur des outils dont on ne comprend pas l'utilité. On part d'un problème. Si on veut comprendre un moteur, on ne commence pas par un cours sur les tournevis. On donne un moteur aux enfants, et on leur demande de le démonter. Les outils viennent naturellement, parce qu'on en a besoin. Et surtout, Musk a une image que je trouve magnifique : il faut apprendre en arbre. D'abord le tronc, les principes fondamentaux. Puis les grosses branches. Et seulement ensuite les feuilles, les détails. Sinon les détails n'ont rien à quoi s'accrocher, et ils tombent. Notre école fait exactement l'inverse. Elle déverse des milliers de feuilles sur des enfants qui n'ont jamais vu le tronc. Puis elle s'étonne qu'ils oublient tout. Et maintenant, l'IA change absolument tout. En 1984, le chercheur Benjamin Bloom fait une découverte qui aurait dû révolutionner l'éducation. Un élève qui bénéficie d'un tuteur particulier progresse de deux écarts-types par rapport à un élève en classe classique. Concrètement, un élève moyen devient meilleur que 98 % de ses camarades. Le problème : on ne pouvait pas offrir un précepteur à chaque enfant. Trop cher. Réservé aux princes. Aristote était le précepteur d'Alexandre le Grand. Pendant 2 500 ans, ce privilège a été réservé aux fils de rois. Aujourd'hui, chaque enfant de la planète peut avoir son Aristote dans sa poche. Un tuteur infiniment patient, disponible à toute heure, qui s'adapte à son rythme, qui connaît toutes les disciplines, et qui peut lui faire remonter l'arbre du savoir du tronc jusqu'aux feuilles. C'est la plus grande révolution éducative depuis l'imprimerie. Et la gauche va mourir dessus. Parce que tout son pouvoir reposait sur le monopole de la transmission. Quand chaque enfant peut apprendre directement, sans passer par l'idéologie de l'institution, le monopole s'effondre. Alors voici ce que je pense qu'il faudrait faire. Que des milliardaires reprennent des universités privées. Il y en a des dizaines en difficulté chaque année. Qu'ils virent toute la bureaucratie administrative qui a proliféré pendant trente ans. Et qu'ils en fassent des écoles de droite, fondées sur le mérite, la compétence et la recherche de la vérité. Pas des écoles qui remplacent une idéologie par une autre. Des écoles qui apprennent à penser, pas quoi penser. Le modèle existe déjà. Il a 2 500 ans. À Athènes, on ne formait pas des spécialistes. On formait des hommes complets. Au Lycée d'Aristote, on étudiait la biologie le matin, la politique l'après-midi, la logique le soir. On marchait en débattant. On cherchait la vérité ensemble, sans peur, sans tabou. Les Grecs avaient un mot pour ça : la paideia. La formation de l'être tout entier. C'est de là que sont sortis ceux qui ont inventé la philosophie, la démocratie, la géométrie, le théâtre, la médecine. En quelques générations, une petite cité a posé les fondations de toute la civilisation occidentale. Voilà ce qu'on doit reconstruire. Des écoles de polymathes. Des esprits libres capables de coder et de lire Platon, de comprendre une équation et de défendre une idée, de douter de tout y compris d'eux-mêmes. Des penseurs qui travailleront avec l'IA comme copilote et qui inventeront des choses que nous ne pouvons même pas imaginer aujourd'hui. Athènes avait des esclaves pour libérer du temps aux penseurs. Nous aurons des machines. Pour la première fois de l'histoire, une civilisation entière peut devenir Athènes. Ce n'est pas la fin de l'école. C'est sa renaissance.

My first sole-author paper was accepted at NeurIPs 🥳🥳🚀